26 August 2022 in Physical Review Fluids
Conventional empirical turbulence modeling is progressive: one begins by modeling simple flows and progressively works towards more complex ones. The outcome is a series of nested models, with the next, more complex model accounting for some additional physics relative to the previous, less complex model. The above, however, is not the philosophy of data-enabled turbulence modeling. Data-enabled modeling is one stop: one trains against a group of data, which contains simple and complex flows. The resulting model is the best fit of the training data but does not closely reproduce any particular flow. The differences between the two modeling approaches have left data-enabled models open to criticism: machine learned models do not fully preserve, e.g., the law of the wall (among other empirical facts), and they do not generalize to, e.g., high Reynolds numbers (among other conditions). The purpose of this paper is to respond to and resolve some of these criticisms: we intend to show that the conventional progressive modeling is compatible with data-enabled modeling. The paper hinges on the extrapolation theorem and the neutral neural network theorem. The extrapolation theorem allows us to control a network's behavior when extrapolating and the neutral neural network theorem allows us to augment a network without “catastrophic forgetting.” For demonstration purposes, we successively model the flow in the constant stress layer, which is simple; the flow in a channel and a boundary layer, which is more complex; and wall-bounded flow with system rotation, which is even more complex. We show that the more complex models respect the less complex models, and that the models preserve the known empiricism.
11 May 2021 in Journal of Fluid Mechanics
The reshocked turbulent Richtmyer–Meshkov (RM) mixing of two media is the most representative problem of more general and complex turbulent mixing induced by interfacial instabilities, broadly occurring in both nature and engineering applications. An accurate prediction of its evolving of spatial structure and mixing width (MW) is of fundamental importance. However, satisfactory prediction with the large-eddy simulation (LES) has not yet been achieved, even for the most important MW. In this paper, we innovatively solve this problem by combining the idea of the constrained large-eddy simulation (CLES), which succeeded previously only in classical single-medium turbulence, and our recently developed Reynolds averaged Navier–Stokes (RANS) model, which realized a satisfactory prediction of MW. Specifically, in our currently developed CLES model, with the aid of Reynolds decomposition, the unclosed subgrid scale (SGS) LES model is decomposed into two parts, i.e. the averaged and the fluctuating. The averaged part is dominated and modelled by the counterpart of our recently developed RANS model to accurately predict the MW, while the fluctuating part is modelled with the classical Smagorinsky model. Consequently, besides successfully capturing the three-dimensional large-scale structure of turbulence and the evolution of the (normalized) mixed mass, our newly proposed CLES also predicts a satisfactory MW with a very coarse grid. To the best of our knowledge, this is the first time that the LES can yield such a comparable result with experiment.
7 February 2024 in Theoretical and Applied Mechanics Letters
14 November 2022 in Journal of Fluid Mechanics
We conduct direct numerical simulations and study the evolution of a pair of counter-rotating vortices in a stratified and turbulent environment beyond the first vortex linking (which is due to the Crow instability). The initial position of the two vortices is perpendicular to the direction of thermal stratification. We vary the Froude number, the background turbulence intensity and the Reynolds number, covering strong, weak and neutral stratification; low, moderate and high background turbulence; and low and moderate Reynolds numbers. We observe a second vortex linking in a weakly stratified environment for the first time. The second vortex linking is followed by turbulence bursts that lead to the rapid decay of the residual vortices after the first vortex linking, leading to a short-living vortex pair. This challenges the conventional view that residual vortices have a long life span, which is true only in an unstratified environment. In addition to the second vortex linking, we observe the following. First, strong background turbulence quickly breaks the two vortices. Second, the two vortices induce secondary and tertiary vortices and lose energy to gravitational waves in a strongly stratified ( $Fr\leqslant 0.7$ ) environment.
21 December 2023 in Physical Review Fluids
Assessing the compliance of a white-box turbulence model with known turbulent knowledge is straightforward. It enables users to screen conventional turbulence models and identify apparent inadequacies, thereby allowing for a more focused and fruitful validation and verification. However, comparing a black-box machine-learning model to known empirical scalings is not straightforward. Unless one implements and tests the model, it would not be clear if a machine-learning model, trained at finite Reynolds numbers preserves the known high Reynolds number limit. This is inconvenient, particularly because model implementation involves retraining and reinterfacing. This work attempts to address this issue, allowing fast a priori screening of machine-learning models that are based on feed-forward neural networks (FNN). The method leverages the mathematical theorems we present in the paper. These theorems offer estimates of a network's limits even when the exact weights and biases are unknown. For demonstration purposes, we screen existing machine-learning wall models and RANS models for their compliance with the log layer physics and the viscous layer physics in an a priori manner. In addition, the theorems serve as essential guidelines for future machine-learning models.
15 October 2025 in Journal of Computational Physics